TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK	REMOVE
Pedestrian Detection	Caltech	NMS-Loss	Reasonable Miss Rate	2.92	# 6
Pedestrian Detection	CityPersons	NMS-Loss	Reasonable MR^-2	10.08	# 10

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/nms-loss-learning-with-non-maximum/pedestrian-detection-on-caltech)](https://paperswithcode.com/sota/pedestrian-detection-on-caltech?p=nms-loss-learning-with-non-maximum)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/nms-loss-learning-with-non-maximum/pedestrian-detection-on-citypersons)](https://paperswithcode.com/sota/pedestrian-detection-on-citypersons?p=nms-loss-learning-with-non-maximum)`

NMS-Loss: Learning with Non-Maximum Suppression for Crowded Pedestrian Detection

4 Jun 2021 · Zekun Luo, Zheng Fang, Sixiao Zheng, Yabiao Wang, Yanwei Fu ·

Non-Maximum Suppression (NMS) is essential for object detection and affects the evaluation results by incorporating False Positives (FP) and False Negatives (FN), especially in crowd occlusion scenes. In this paper, we raise the problem of weak connection between the training targets and the evaluation metrics caused by NMS and propose a novel NMS-Loss making the NMS procedure can be trained end-to-end without any additional network parameters. Our NMS-Loss punishes two cases when FP is not suppressed and FN is wrongly eliminated by NMS. Specifically, we propose a pull loss to pull predictions with the same target close to each other, and a push loss to push predictions with different targets away from each other. Experimental results show that with the help of NMS-Loss, our detector, namely NMS-Ped, achieves impressive results with Miss Rate of 5.92% on Caltech dataset and 10.08% on CityPersons dataset, which are both better than state-of-the-art competitors.

PDF Abstract